HA(high availability) ETA(estimated time of arrival) caching service and cluster
- docker with docker-compose
- golang (for development)
Architecture overview:
+-------------+ +----------+
| | | |
| nginx(1s) | | cache | +-------------+
| | | | | |
+---->+ cache-01 +-------+ node-01 +--------+ |
+------------------+ | | | | | | |
| | | +-------------+ +----------+ | |
+------+ balancer +-------------+ | |
| | | | +-------------+ +-----------+ | |
| +------------------+ | | | | | | |
| round-robin +---->+ cache-02 +------>+ node-02 +-------> GEO API |
| +------------------+ | | | | | | |
| | | | +-------------+ +-----------+ | |
+----->+ balancer.backup | | | |
| | | +-------------+ +-----------+ | |
+------------------+ | | | | | | |
+---->+ cache-03 +------>+ node-03 +------>+ |
VIP / keepalived | | | | | |
+-------------+ +-----------+ +-------------+
- balancer: via nginx round-robin upstream balancing. HAProxy can also be used;
- cache-0[1-3]: simple http 200 cache with 1s ttl;
- node-0[1-3]: worker nodes;
- GEO API: API for geo location and prediction;
Clone this app with:
$ git clone https://github.com/vany-egorov/ha-eta.gitRun under docker (docker + docker-compose required). This will build app, run tests, start docker-containers. Cluster listen on 9179 port
$ cd ha-eta
$ chmod +x ./run.sh
$ ./run.shapp can simply be tested with:
$ curl -XGET 'http://127.0.0.1:9179/api/v1/eta/min?lat=55.752992&lng=37.618333'
# or
$ watch -n -1 "curl -XGET 'http://127.0.0.1:9179/api/v1/eta/min?lat=55.752992&lng=37.618333'"build app and run without docker (golang required) for development and local run and testing:
$ cd ha-eta/src
$ chmod +x ./build.sh
$ ./build.sh
$ ./ha-eta --help
# сustom port
$ ./ha-eta -p 8888
# disable cache and custom port
$ ./ha-eta -p 8888 --do-not-cache-points --do-not-cache-etas
# increase cache TTL
$ ./ha-eta -p 8888 --cache-points-ttl 15s --cache-etas-ttl 30m
# custom GEO engine url
$ ./ha-eta -p 8888 --cache-points-ttl 15s --cache-etas-ttl 30m --wheely-url https://prod-api.wheely.com/eta